The tools rewriting the threat model
Major AI
The frontier labs and the platforms built on them. Each hub collects our coverage — news, management changes, and releases.
OpenAI
The company whose November 2022 chatbot launch turned a research field into a consumer market — and whose releases have set the industry's pace, and its security questions, ever since.
Google DeepMind
The research powerhouse that invented the transformer and then found itself racing its own students — from AlphaFold's scientific triumphs to the Gemini era.
Anthropic
The safety-focused lab behind the Claude model family — and, for this magazine's beat, the company whose disclosures about the misuse of AI agents have repeatedly defined how defenders think about machine-speed attackers.
- From The VaultNovember 2025 — The Attacker Becomes an Algorithm
Microsoft
The distribution machine: the company that put a frontier model inside Windows, Office and a security product line — and tied its fortunes to OpenAI with the biggest cheque in the industry.
- From The VaultJune 2022 — GitHub Copilot goes on general sale
Meta AI
The open-weights standard-bearer. Meta's releases changed who gets to run a capable model — and its stumbles, like the three-day life of Galactica, changed how labs launch.
- From The VaultNovember 2022 — Galactica, withdrawn in 72 hours
xAI & the challengers
The fast followers: the labs proving that frontier capability no longer belongs to three companies — and forcing the pricing, openness and safety debates the incumbents would rather schedule themselves.
- Coverage buildingArticles land here as they publish.
Minor tools & the long tail
Image generators, coding assistants, voice models, agent frameworks — the small launches that quietly change workflows. These are tracked in the monthly AI Tech desk of every Vault edition rather than given hubs of their own — Midjourney and Stable Diffusion in the summer of 2022, Copilot and the coding assistants before them, and each month's notable releases across the whole archive.
The monthly AI Tech desk
Every Vault edition carries an AI Tech desk — the month's launches of major and minor AI tools, management moves and releases, read with hindsight. The run is unbroken across all one hundred and fifty-eight editions: 2014 opens with Google buying DeepMind and closes with Hawking's warning and the debut of generative adversarial networks; 2015 runs from the Atari agent in February to TensorFlow in November and OpenAI's founding in December; 2016 opens with AlphaGo's defeat of Lee Sedol, a chatbot withdrawn within a day, and the first death of a driver under Autopilot; 2018 brings BERT and the first pedestrian killed by a self-driving car; 2019 brings AlphaStar and GPT-2 withheld as too dangerous to release; 2020 brings Clearview AI and the GPT-3 paper; 2022 runs DALL·E 2 to Stable Diffusion to ChatGPT; 2023 carries GPT-4, Claude and the OpenAI board weekend; and the desks continue month by month to August 2026.
The Sentry Briefing
AI launches and the security news around them — once a week, in your inbox.